Later chronotype is associated with higher hemoglobin A1c in prediabetes patients

ABSTRACT

The circadian system is known to play a role in glucose metabolism. Chronotype reflects the interindividual variability in the phase of entrainment. Those with later chronotype typically prefer later times in the day for different activities such as sleep or meals. Later chronotype has been shown to be associated with metabolic syndrome, increased diabetes risk and poorer glycemic control in type 2 diabetes patients. In addition, “social jetlag”, a form of circadian misalignment due to a mismatch between social rhythms and the circadian clock, has been shown to be associated with insulin resistance. Other sleep disturbances (insufficient sleep, poor sleep quality and sleep apnea) have also been shown to affect glucose metabolism. In this study, we explored whether there was a relationship between chronotype, social jetlag and hemoglobin A1c (HbA1c) levels in prediabetes patients, independent of other sleep disturbances. A cross-sectional study was conducted at the Department of Family Medicine, Ramathibodi Hospital, Bangkok, from October 2014 to March 2016 in 1014 non-shift working adults with prediabetes. Mid-sleep time on free day adjusted for sleep debt (MSFsc) was used as an indicator of chronotype. Social jetlag was calculated based on the absolute difference between mid-sleep time on weekdays and weekends. The most recent HbA1c values and lipid levels were retrieved from clinical laboratory databases. Univariate analyses revealed that later MSFsc (p = 0.028) but not social jetlag (p = 0.48) was significantly associated with higher HbA1c levels. Multivariate linear regression analysis was applied to determine whether an independent association between MSFsc and HbA1c level existed. After adjusting for age, sex, alcohol use, body mass index (BMI), social jetlag, sleep duration, sleep quality and sleep apnea risk, later MSFsc was significantly associated with higher HbA1c level (B = 0.019, 95% CI: 0.00001, 0.038, p = 0.049). The effect size of one hour later MSFsc on HbA1c (standardized coefficient = 0.065) was approximately 74% of that of the effect of one unit (kg/m²) increase in BMI (standardized coefficient = 0.087). In summary, later chronotype is associated with higher HbA1c levels in patients with prediabetes, independent of social jetlag and other sleep disturbances. Further research regarding the potential role of chronotype in diabetes prevention should be explored.

KEYWORDS:

• Chronotype
• circadian rhythm
• hemoglobin A1c
• prediabetes
• sleep

Introduction

In the United States, the prevalence of diabetes was estimated to be 9.3% in 2012 (National Center for Chronic Disease Prevention and Health Promotion, 2014). Three times as many, 37%, were estimated to have prediabetes based on hemoglobin A1c (HbA1c) or fasting glucose levels, translating to 86 million people. Globally, an estimated 318 million people had impaired glucose tolerance in 2015 according to the International Diabetes Federation, with a projected alarming increase to 482 million in 2040. Without interventions, 5.8%–18.3% of people with prediabetes develop diabetes yearly (Knowler et al., 2002; Ramachandran et al., 2006; Tuomilehto et al., 2001). Although some of the diabetes risk factors, such as family history or ethnicity, are not modifiable, lifestyle interventions with diet and exercise can reduce diabetes risk by as much as 58% (Knowler et al., 2002; Tuomilehto et al., 2001). Therefore, identifying additional modifiable risk factors may lead to a research to develop novel lifestyle interventions to reduce incident diabetes.

The circadian system, controlled by the master circadian clock located in the hypothalamus, plays a major role in regulating daily rhythms of sleep/wake cycle, central and peripheral tissue metabolism, and hormonal secretions (Huang et al., 2011). The central clock is synchronized to the light-dark cycle and relays the information via various pathways to the peripheral organs, leading to coordinated rhythms. Circadian misalignment occurs when sleep and/or meal timing is out of synchrony with the light-dark cycle (environment) or the central circadian clock (endogenous), or when there is a desynchrony between the central clock and the peripheral clocks in peripheral organs. There is evidence that the circadian system plays a role in glucose metabolism and that circadian misalignment can lead to glucose intolerance. Experimental circadian misalignment in healthy humans resulted in reduced glucose tolerance and increased inflammatory markers, suggesting an increase in cardiometabolic risk (Buxton et al., 2012; Leproult et al., 2014; McHill et al., 2014; Morris et al., 2015; Scheer et al., 2009). In our modern society, night-shift work, which requires workers to eat and be active during their circadian night and sleep during their circadian day, represents an extreme form of circadian misalignment. Night-shift workers have increased risk of health problems, including obesity (Pan et al., 2011), metabolic syndrome (Karlsson et al., 2003) and cardiovascular disease (Boggild et al., 1999). Moreover, a recent meta-analysis of 10 cohort studies (262 294 participants) revealed that shift work was associated with a 40% increase in the risk of developing diabetes (Anothaisintawee et al., 2015). Another meta-analysis also reported an increased diabetes risk in shift workers, although at a lower estimate of 9% (Gan et al., 2015). There appeared to be an increasing risk in those with a longer duration of shift work (Pan et al., 2011).

At an individual level, the circadian clock entrains differently to the light-dark zeitgeber, earlier or later depending on the characteristics of the clock. The relationship between the internal day and external day (e.g. between the minimum of the core body temperature and dawn) is called “phase of entrainment” or “chronotype” (Roenneberg & Merrow, 2016). Chronotype is also used to describe personality traits associated with the preferred times of the day for activities such as sleep or meal intake (Roenneberg & Merrow, 2016). Evening types (or late chronotypes) typically have later bedtime than those who are morning types (or early chronotypes). Questionnaires have been developed to assess morningness-eveningness preference (Horne & Ostberg, 1976; Smith et al., 1989). In addition, chronotype can be assessed by using the mid-sleep times (midpoint of sleep between sleep onset and wake time) on free days, with further correction for sleep debt accumulated over the work week (herein will be referred to as MSFsc) and can be captured using the Munich ChronoType Questionnaire (Roenneberg et al., 2007b). Those with later chronotype usually experience a mild form of circadian misalignment due to a greater degree of misalignment between social rhythms and the circadian clock, a phenomenon called “social jetlag” (Wittmann et al., 2006). Social jetlag results from shifting sleep timing between workdays and free days resembling traveling across time zones. In non-shift workers, recent evidence suggested that evening preference or later chronotype as well as greater social jetlag was associated with adverse effects on cardiometabolic function, including higher body mass index (BMI) or overweight/obesity (Olds et al., 2011; Parsons et al., 2015; Roenneberg et al., 2012) and metabolic syndrome (Parsons et al., 2015; Yu et al., 2015). In some of these studies that measured chronotype and social jetlag simultaneously, the associations between chronotype and metabolic parameters were not significant after adjusting for social jetlag (Parsons et al., 2015; Roenneberg et al., 2012). Recently, two large population-based studies of more than 6000 participants revealed that evening chronotype was associated with an increased risk of having type 2 diabetes [odds ratio 1.73 (Yu et al., 2015) and 2.5 in men and women combined (Merikanto et al., 2013)]. However, in the Nurses’ Health Study 2 involving 64 615 women followed for 6 years, later chronotypes were not a risk factor for type 2 diabetes while early chronotypes had a 13% reduction in risk compared to intermediate chronotypes (Vetter et al., 2015). Among patients with type 2 diabetes, later chronotypes and evening preferences have been found to be associated with poorer glycemic control (Osonoi et al., 2014; Reutrakul et al., 2013). These data demonstrate the contribution of circadian regulation on glucose metabolism.

In addition to circadian regulation, other sleep disturbances are known to influence glucose metabolism, including poor sleep quality, abnormal sleep duration and obstructive sleep apnea (OSA) (Reutrakul & Van Cauter, 2014). To date, little is known about the role of circadian regulation on glucose metabolism in those with prediabetes or impaired glucose metabolism. Since this is a high-risk population for developing diabetes, targeted interventions could be further studied if such a relationship is found. The aim of this study was to examine whether chronotype, as assessed by MSFsc, and social jetlag were independently associated with HbA1c levels in patients with prediabetes. We hypothesized that later chronotype would be associated with higher HbA1c levels, independent of other sleep disturbances and relevant covariates.

Materials and methods

This cross-sectional study is a part of a prediabetes (PreDM) cohort study. In brief, this cohort study was conducted at the outpatient clinic of Department of Family Medicine, Ramathibodi Hospital, Bangkok, Thailand, and the participants were recruited during October 2014 to March 2016. Adult patients with a diagnosis of prediabetes, defined as fasting plasma glucose (FPG) between 100 and 125 mg/dl (5.6 and 6.9 mmol/L) or HbA1c between 5.70% and 6.49% (38.80–47.44 mmol/mol), were invited to participate (American Diabetes Association, 2016). The participants will be followed for at least 5 years or until they develop diabetes mellitus. Baseline information was used for this cross-sectional study. Participants were excluded if they were shift workers, or had HbA1c level ≥6.5% (48.0 mmol/mol) or FPG ≥126 mg/dl (≥7.0 mmol/L). The study’s protocol was approved by the Ethical Clearance Committee, Faculty of Medicine Ramathibodi Hospital. All participants gave written informed consent.

Data collection

Participants were interviewed to obtain information regarding age, sex, marital status, educational level (primary school, secondary school or college), family history of diabetes mellitus in first-degree relatives, history of smoking (never versus current/past users) and alcohol use (never versus current/past users). Date of diagnosis of prediabetes, history of underlying diseases (i.e. diagnosis of hypertension or dyslipidemia) and height were extracted from the patient’s medical records by investigating physicians (TA, ST and DL). Weight was measured on the date of interview. BMI was calculated by dividing weight (kilogram) by height² (meter²).

In addition, depressive symptoms, previously reported to be related to glycemic control (Lustman et al., 2000), were assessed using the validated Thai version of the Center for Epidemiologic Studies-Depression (CES-D) scale (Radloff, 1977; Trangkasombat et al., 1997). The scores ranged from 0 to 60, of which higher scores indicate more severe depressive symptoms.

Chronotype and social jetlag assessments

Participants reported their usual bedtime, wake-up time, sleep onset latency and actual sleep duration on weekdays and weekends over the previous month. From these, we calculated the mid-sleep time separately for weekdays and weekends as the midpoint between sleep onset and wake time. The metric of chronotype, mid-sleep time on free days adjusting for sleep debt (MSFsc), was derived from mid-sleep time on weekend nights with further adjustment for the sleep debt taking into account the sleep duration average of weekends and weekdays as follows: MSFsc = mid-sleep time on weekend night – 0.5*[SD_F -(5*SD_W + 2*SD_F)/7], where SD_F is the calculated sleep duration on weekend nights and SD_w is the calculated sleep duration on weekday nights, as outlined in the Munich ChronoType Questionnaire (Roenneberg et al., 2004, 2007a]. Social jetlag was calculated based on the absolute difference between mid-sleep time on weekdays and weekends (Roenneberg et al., 2012).

Sleep assessments

Average self-reported sleep duration was calculated as [(sleep duration on weekdays * 5) + (actual sleep duration on weekend * 2)]/7. Sleep duration was derived from the question “During the past month, how many hours of actual sleep did you get at night?”, which was asked separately for weekdays and weekends.

To assess sleep quality independent of sleep duration, we utilized a modified Pittsburgh Sleep Quality Index (PSQI) score (Knutson et al., 2006b). The PSQI score evaluates sleep duration and quality within the past month, with a higher score indicating worse sleep (Buysse et al., 1989), and has been validated in a Thai population (Sitasuwan et al., 2014). The modified PSQI score excludes the sleep duration component from the PSQI (Knutson et al., 2006b).

Participants reported whether they had a diagnosis of OSA. Those without a previous diagnosis were interviewed using the Berlin questionnaire to assess the risk of having OSA, which categorizes respondents as high or low risk of having OSA (Netzer et al., 1999). The questionnaire was previously validated in a Thai population (Suksakorn et al., 2014). Participants who had a diagnosis of or were at high risk for OSA were grouped together as the presence or high risk of OSA (OSA risk).

HbA1c and lipid levels

HbA1c reflects an average of glucose levels in the preceding three months. The most recent HbA1c values of study’s participants were retrieved from laboratory databases, Medical statistic Unit, Ramathibodi Hospital. Around 43% of the HbA1c values were obtained on the date of the interview, 21.3% were obtained before and 35.5% after the date of the interview (at an average of 3 and 4 months), respectively. The time lag between the date of performing HbA1c and the date of interview did not exceed 180 days. The HbA1c assay at Ramathibodi Hospital has been NGSP (National Glycohemoglobin Standardization Program) certified.

Lipid levels, within one year of the interview date, were obtained from medical records. These included total cholesterol, high-density lipoprotein (HDL), low-density lipoprotein (LDL) and triglycerides levels. A one -year interval was chosen as the American Heart Association recommended monitoring lipid levels every 3–12 months in those on treatment as clinically indicated (Stone et al., 2014)

Statistical analysis

Normal distribution of study data was evaluated by exploring skewness and kurtosis of the data. The data were presented as means and standard deviations (SDs), if the data were normally distributed; otherwise, they were presented as medians and interquartile ranges (IQRs). Univariate linear regression analysis was applied to assess the association among demographic (i.e. age, sex, educational level, history of smoking and alcohol use, BMI, depressive symptoms and lipid levels), sleep parameters, (i.e. sleep quality, sleep duration and OSA risk), MSFsc and social jetlag, and HbA1c level. Multivariate linear regression analysis was applied to determine the independent association between MSFsc and HbA1c levels. Variables that had p-values less than 0.10 from the univariate linear regression model were considered in the multivariate linear regression analysis. In addition, age, sleep duration, sleep quality, OSA risk and social jetlag were included as they were previously reported to be associated with glycemic control (Knutson et al., 2006b; Parsons et al., 2015; Reutrakul & Van Cauter, 2014). Lastly, interactions between MSFsc and different variables, including sex, BMI, OSA risk, sleep duration and sleep quality, were assessed.

p-values less than 0.05 were considered significant. All analyses were performed using STATA version 14.

Results

A total of 1014 participants were included in the study. Their baseline demographic, circadian and sleep characteristics are presented in Table 1. The mean age of participants was 62.4 (SD = 8.7) years. Two-thirds of the participants were female (66.5%) and had been diagnosed with hypertension (68.4%). Mean BMI was 26.0 (4.0) kg/m² and most of the study’s participants had a diagnosis of dyslipidemia as an underlying disease. Median HbA1c value (IQR) was 5.83% (4.31%–6.49%) [40.2 mmol/mol, (23.6–47.4)] and median CES-D score was 7 (0–46). On average, participants had slightly later bedtimes and wake times on weekends than on weekdays with a mean MSFsc of 1:57 (1:11) a.m. Mean self-reported sleep duration was 5.80 (1.45) hours and OSA risk was found in approximately 30% of the participants. None of the participants was using metformin.

Table 1. Descriptive demographic, sleep and circadian parameters (n = 1014).

Demographic data
Age (years)	62.4 (8.7)
Female	674 (66.5)
Educational level	361 (35.7)
Primary school or less	274 (27.1)
Secondary school	377 (37.3)
College or higher
Smoking status	966 (95.3)
Never Current/past	48 (4.7)
Alcohol use
Never	822 (81.1)
Current/past	192 (18.9)
Family history of diabetes	435 (42.9)
Hypertension	691 (68.4)
Dyslipidemia	922 (91.1)
Total Cholesterol (mg/dL) High-density lipoprotein(mg/dL)	206.1 (36.0) 54.5 (13.9) 126.2 (31.6)
Low-density lipoprotein (mg/dL)	139.9 (71.3)
Triglycerides (mg/dL)
BMI (kg/m^2)	26.0 (4.0)
HbA1c (%)	5.83 (4.31-6.49)
HbA1c (mmol/mol)	40.2 (23.6-47.4)
CES-D score	7 (0-46)
Chronotype and social jetlag
Chronotype (MSFsc, h:min)	01:57 (1:11)
Bedtime weekday (h:min)	21:58 (1:15)
Bedtime weekend (h:min)	22:03 (1:18)
Wake time weekday (h:min)	5:15 (1:13)
Wake time weekend (h:min)	5:31 (1:21)
Social jetlag (min)	0 (0–192)
Sleep parameters
Sleep duration (h)	5.80 (1.45)
Modified PSQI score	4 (0–15)
OSA risk	303 (29.9)

Data are presented as mean (SD), median (inter-quartile range) or n (%).

Time is presented in 24-hour clock time

Association between HbA1c and demographics, sleep and circadian parameters

To explore the correlations between HbA1c levels and demographics, sleep and circadian parameters, univariate linear regression analyses were performed (Table 2). Age, educational level, smoking and depressive symptoms were not significantly associated with HbA1c levels. Being female and having higher BMI were associated with higher HbA1c levels while those consuming alcohol had lower HbA1c when compared with nonusers. Triglycerides, but not other lipid levels, were associated with higher HbA1c. Later chronotype (MSFsc) was found to be significantly associated with higher HbA1c [β-coefficient = 0.20 (95% CI: 0.002, 0.038), p = 0.028], while social jetlag was not significantly related with HbA1c. For sleep parameters, participants with a high risk of OSA had significantly higher HbA1c levels than those with a low risk of OSA. Sleep duration and sleep quality were not found to be associated with HbA1c levels.

Table 2. Univariate linear regression analysis between HbA1c and demographic, sleep and circadian parameters.

	Unstandardized coefficient (B)	95% CI for B	Standardized coefficient	p-value
Demographic data
Age (years)	−0.001	−0.004, 0.001	−0.031	0.326
Female	0.09	0.04, 0.13	0.120	<0.001
Educational level^a
Secondary school	0.01	−0.04, 0.06	0.014	0.705
College or higher	−0.001	−0.051, 0.048	−0.002	0.964
Current/past smoking	−0.03	−0.13, 0.07	−0.020	0.508
Alcohol use	−0.06	−0.12, −0.01	−0.070	0.023
Cholesterol (mg/dL)	0.0002	−0.0004, 0.0008	0.018	0.580
HDL (mg/dL)	−0.001	−0.003, 0.0004	−0.049	0.131
LDL (mg/dL)	0.0002	−0.003, 0.0004	0.017	0.586
Triglycerides (mg/dL)	0.0004	0.0001, 0.0007	0.087	0.007
BMI	0.011	0.005, 0.016	0.123	<0.001
CES-D score	0.0001	−0.004, 0.004	0.0021	0.947
Chronotype and social Jetlag
Chronotype (MSFsc)	0.020	0.002, 0.038	0.069	0.028
Social jet lag (min)	0.02	−0.03, 0.06	0.020	0.484
Sleep parameters
Sleep duration	−0.01	−0.02, 0.01	−0.025	0.420
Modified PSQI	−0.0004	−0.008, 0.007	−0.004	0.911
OSA risk	0.08	0.03, 0.12	0.103	0.001

^a Reference = primary school or less.

Multiple regression analysis was performed to determine whether chronotype was independently associated with HbA1c (Table 3, n = 960). After adjusting for age, sex, alcohol use, BMI, triglycerides, sleep duration, sleep quality (modified PSQI), OSA risk and social jetlag, later MSFsc remained significantly associated with higher HbA1c level (B = 0.019, 95% CI: 0.00001, 0.038, p = 0.049). In addition, female sex (p = 0.001), higher BMI (p = 0.009), higher triglycerides levels (p = 0.007) and being at high risk for OSA (p = 0.004) were independently associated with a higher HbA1c level. There were no significant interactions found between MSFsc and different variables, including sex, BMI (<25 versus ≥25 kg/m²), OSA risk (present or absent), sleep duration (<7, 7–8, >8 h) and sleep quality.

Table 3. Multiple regression analysis with HbA1c as an outcome (n = 960).

	Unstandardized coefficient (B)	95% CI for B	Standardized coefficient	95% CI for standardized coefficient	p-value
Age (years)	−0.0005	−0.003, 0.002	−0.012	−0.079, 0.054	0.721
Female	0.082	0.034, 0.130	0.114	0.048, 0.180	0.001
Alcohol use	−0.065	−0.124, −0.007	−0.075	−0.142, −0.008	0.029
BMI (kg/m^2)	0.007	0.002, 0.013	0.087	0.021, 0.152	0.009
Triglyceride (mg/dl)	0.0004	0.0001, 0.0007	0.086	0.023, 0.149	0.007
Chronotype (MSFsc)	0.019	0.00001, 0.038	0.065	0.0004, 0.130	0.049
Social jetlag (min)	0.018	−0.028, 0.064	0.026	−0.039, 0.090	0.435
Sleep duration (h)	−0.008	−0.026, 0.010	−0.033	−0.112, 0.045	0.405
Modified PSQI	−0.007	−0.017, 0.002	−0.060	−0.139, 0.020	0.141
OSA risk	0.071	0.023, 0.119	0.096	0.032, 0.160	0.004

Discussion

In this large cohort, we demonstrated for the first time that later chronotype is independently associated with higher HbA1c levels in patients with prediabetes, after adjusting for demographics and sleep disturbances, including sleep duration, sleep quality and the risk for OSA, as well as social jetlag. For example, two individuals with prediabetes whose MSFsc differs by two hours (all other covariates the same) would be expected to have different HbA1c levels, specifically 5.70% (38.8 mmol/mol) versus 5.74% (39.2 mmol/mol). Although this effect size is relatively modest, it is approximately 74% of the effect of 1 unit of BMI on HbA1c levels in this cohort (as indicated by standardized coefficients, Table 3), and BMI is one of the strongest risk factors for diabetes. It is possible that the relatively small effect size, compared to the previous report in patients with type 2 diabetes (Reutrakul et al., 2013), is related to the narrower range of HbA1c values in this population. Although later MSFsc is associated with more social jetlag in this study (data not shown), social jetlag itself was not related to glycemic status. These data further support the role of circadian regulation on glucose metabolism and raise the possibility of sleep timing adjustment as a diabetes prevention strategy.

Studies experimentally manipulating levels of circadian misalignment in healthy volunteers have elucidated possible mechanisms linking circadian misalignment to abnormal glucose metabolism. Increased glucose levels (both fasting and postprandial, between 6% and 17%), without adequate pancreatic β-cell insulin response, were found after 6–21 days of circadian misalignment (Buxton et al., 2012; Leproult et al., 2014; McHill et al., 2014; Morris et al., 2015; Scheer et al., 2009). This was accompanied by a worsening of cardiometabolic parameters including increased mean arterial blood pressure (Scheer et al., 2009), decreased energy expenditure (McHill et al., 2014), elevated inflammatory markers (Leproult et al., 2014) and free fatty acids (Morris et al., 2015), and alterations in appetite-regulating hormones (Scheer et al., 2009). These changes were found to be independent of changes in sleep duration (Leproult et al., 2014). In a population-based study, evening chronotype, which is typically associated with a mild form of circadian misalignment, has been associated with metabolic syndrome (Yu et al., 2015), type 2 diabetes (Merikanto et al., 2013; Yu et al., 2015) and lower HDL cholesterol (Wong et al., 2015).

In our study, social jetlag, one indicator of circadian misalignment, was not related with HbA1c level. Whether this was related to the fact that the majority of our participants (68%) had no social jetlag was uncertain. However, this suggests that the association between later chronotype and glucose metabolism is independent of social jetlag in our sample. The effect of circadian phase on glucose metabolism may be different from that of circadian misalignment. In an experimental study designed to distinguish the effects of the endogenous circadian system and circadian misalignment, the circadian system and circadian misalignment had independent influences on glucose metabolism (Morris et al., 2015). Postprandial glucose levels were 17% higher in the biological evening than in the morning. The early-phase postprandial insulin response was 27% lower in the evening, indicative of insufficient β-cell response influenced by circadian phase, while circadian misalignment increased the postprandial glucose levels by 6% despite a 14% higher late-phase postprandial insulin response, suggesting reduced insulin sensitivity (Morris et al., 2015).

As later chronotypes typically sleep and eat later during the 24 h day, these behaviors may be potential mediators linking chronotype to cardiometabolic risk factors (Reutrakul & Knutson, 2015). Shorter sleep duration was found to be associated with later chronotype (Fabbian et al., 2016), possibly partly due to earlier wake time than desired, especially during workdays, to conform to normal social schedule. In addition, sleep quality is usually poorer in those with later chronotype (Fabbian et al., 2016). Both short sleep duration and poor sleep quality have been linked to insulin resistance, glucose intolerance and increased diabetes risk (Reutrakul & Van Cauter, 2014). Although later MSFsc was associated with poorer sleep quality in this study (result not shown), sleep duration and quality were not predictors of HbA1c, suggesting other potential mechanisms in this sample. Meal timing can also play a role as exposure to food at an inappropriate time of the day could lead to misalignment between the central and peripheral clocks, resulting in abnormal metabolism and weight gain (Garaulet & Gomez-Abellan, 2014). An experiment in healthy volunteers revealed that isocaloric diet consumption at dinner was found to be associated with an 8% higher postprandial glucose and 14% lower insulin response compared with breakfast time (Morris et al., 2015). Night eating was also reported to be associated with poorer glycemic control in the type 2 diabetes population (Hood et al., 2014). Moreover, a recent randomized crossover study in type 2 diabetes patients found that consuming two larger meals earlier in the day (breakfast and lunch) compared with six small meals for 12 weeks resulted in a greater reduction in body weight and FPG, along with higher insulin sensitivity (Kahleova et al., 2014). Lastly, exposure to artificial light at night could lead to circadian misalignment and altered metabolism. Mice exposed to dim light at night had increased weight gain and reduced glucose tolerance despite an equivalent caloric intake to mice kept in a standard light/dark cycle (Fonken et al., 2010). Nocturnal short-wavelength light, emitted from some electronic devices, was shown to suppress metabolism the following morning in healthy volunteers (Kayaba et al., 2014). Melatonin, a neurohormone secreted by the pineal gland that plays a role in circadian physiology, is suppressed by light. Melatonin receptors are found in the pancreatic β-cell and may modulate insulin secretion (Peschke et al., 2015). Indeed, low nocturnal melatonin secretion was associated with an increased risk of incident diabetes in a large population-based study (McMullan et al., 2013). Collectively, these changes can help explain the association between chronotype and impaired glucose metabolism. Our study is limited by the lack of information on light exposure and meal timing.

Our finding of the independent association between OSA risk and higher HbA1c is in agreement with previous studies as OSA is known to be associated with insulin resistance and contributes to poorer glycemic control, independent of obesity (Pamidi & Tasali, 2012). The effect size of chronotype on HbA1c is approximately two-thirds of that of OSA risk in the current study. Both chronotype and OSA risk independently contributed to higher HbA1c in our study.

The strength of this study is the inclusion of a relatively large number of participants and the use of comprehensive standardized questionnaires. To our knowledge, this is the first study exploring the relationship among chronotype, social jetlag and glycemia in the prediabetes population. However, it has some limitations. Sleep characteristics were not objectively measured. In some participants, HbA1c and sleep assessments did not occur on the same date. However, it has been shown that the results of PSQI remained relatively stable over 1 year (Knutson et al., 2006a). HbA1c itself reflects an average glucose level over three months, and is unlikely to change significantly over a few months in our participants who were not taking any diabetes medications. In addition, information on light exposure at night, meal timing, caloric intake and physical activity were not available. Ideally, including nondiabetic participants would allow us to better understand the relationship between chronotype and glycemia, and explore HbA1c at a broader level. Lastly, the participants were recruited from one center in Thailand, which might limit its generalizability to other regions of Thailand or to other countries.

In summary, later chronotype is independently associated with higher HbA1c levels in patients with prediabetes, supporting the role of circadian regulation in glucose metabolism. Future diabetes prevention research should test the effect of adjusting sleep timing, potentially along with adjusting other related behaviors, including meal timing and artificial light exposure, particularly in a high-risk group with prediabetes.

Declaration of interest

T.A., D.L., S.T., K.L.K. and A.T. have nothing to disclose. S.R. received lecture fees from Sanofi Aventis, Medtronic and Novo Nordisk, equipment support from ResMed, Thailand, and research grant from Merck Sharp & Dohme Corp, U.S.A.

The study was supported in part by a research grant from Investigator-Initiated Studies Program of Merck Sharp & Dohme Corp, U.S.A. (MSIP# 0000-349) and from Thailand Research Organization Network (grant No. 58-051). The opinions expressed in this paper are those of the authors and do not necessarily represent those of Merck Sharp & Dohme Corp.

Acknowledgments

We would like to thank all the participants in the study.

Additional information

Notes on contributors

Thunyarat Anothaisintawee

T.A. conceptualized the study, researched the data, wrote the manuscript, contributed to discussion, reviewed/edited manuscript and is the guarantor of this work and, as such, had full access to the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analyses. D.L. and S.T. researched data and reviewed/edited the manuscript. K.L.K. contributed to discussion, reviewed/edited the manuscript. A.T. analyzed the data, wrote the manuscript, contributed to discussion and reviewed/edited the manuscript. S.R. conceptualized the study, researched the data, wrote the manuscript, contributed to discussion and reviewed/edited the manuscript.

Dumrongrat Lertrattananon

T.A. conceptualized the study, researched the data, wrote the manuscript, contributed to discussion, reviewed/edited manuscript and is the guarantor of this work and, as such, had full access to the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analyses. D.L. and S.T. researched data and reviewed/edited the manuscript. K.L.K. contributed to discussion, reviewed/edited the manuscript. A.T. analyzed the data, wrote the manuscript, contributed to discussion and reviewed/edited the manuscript. S.R. conceptualized the study, researched the data, wrote the manuscript, contributed to discussion and reviewed/edited the manuscript.

Sangsulee Thamakaison

T.A. conceptualized the study, researched the data, wrote the manuscript, contributed to discussion, reviewed/edited manuscript and is the guarantor of this work and, as such, had full access to the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analyses. D.L. and S.T. researched data and reviewed/edited the manuscript. K.L.K. contributed to discussion, reviewed/edited the manuscript. A.T. analyzed the data, wrote the manuscript, contributed to discussion and reviewed/edited the manuscript. S.R. conceptualized the study, researched the data, wrote the manuscript, contributed to discussion and reviewed/edited the manuscript.

Kristen L. Knutson

T.A. conceptualized the study, researched the data, wrote the manuscript, contributed to discussion, reviewed/edited manuscript and is the guarantor of this work and, as such, had full access to the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analyses. D.L. and S.T. researched data and reviewed/edited the manuscript. K.L.K. contributed to discussion, reviewed/edited the manuscript. A.T. analyzed the data, wrote the manuscript, contributed to discussion and reviewed/edited the manuscript. S.R. conceptualized the study, researched the data, wrote the manuscript, contributed to discussion and reviewed/edited the manuscript.

Ammarin Thakkinstian

T.A. conceptualized the study, researched the data, wrote the manuscript, contributed to discussion, reviewed/edited manuscript and is the guarantor of this work and, as such, had full access to the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analyses. D.L. and S.T. researched data and reviewed/edited the manuscript. K.L.K. contributed to discussion, reviewed/edited the manuscript. A.T. analyzed the data, wrote the manuscript, contributed to discussion and reviewed/edited the manuscript. S.R. conceptualized the study, researched the data, wrote the manuscript, contributed to discussion and reviewed/edited the manuscript.

Sirimon Reutrakul

T.A. conceptualized the study, researched the data, wrote the manuscript, contributed to discussion, reviewed/edited manuscript and is the guarantor of this work and, as such, had full access to the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analyses. D.L. and S.T. researched data and reviewed/edited the manuscript. K.L.K. contributed to discussion, reviewed/edited the manuscript. A.T. analyzed the data, wrote the manuscript, contributed to discussion and reviewed/edited the manuscript. S.R. conceptualized the study, researched the data, wrote the manuscript, contributed to discussion and reviewed/edited the manuscript.